In a posting from about a year ago (http://www.digitalmars.com/d/archives/digitalmars/D/learn/842.html), the implicit variable "length" was scheduled for potential
deprecation.

Is it going to be removed? If so, when?

I see this as a significant flaw in the D language. For example, what does the following code do?

  import std.stdio;  // for writefln

  void main() {
    static int[4] foo = [ 0, 1, 2, 3 ];
    int[] bar;
    int length = 2;
    bar = foo[0 .. length];
    writefln("bar = %s", bar);
  }


What is does the program print?
  A. bar = [0,1,2,3]
  B. bar = [0,1]

I would guess that most programmers unfamilar with D would answer "B". However, "A" is the correct answer. It certainly has been a lesson I've had to learn the hard way.

Besides, is "foo[0 .. length]" really better than "foo[0 .. foo.length]"? The first is slightly more convenient but the second is more explicit and readable.

Thanks,
  Bradley

"Bradley Smith" <digitalmars-com@baysmith.com> wrote in message news:e64tgp$1gvu$1@digitaldaemon.com...

> I would guess that most programmers unfamilar with D would answer "B". However, "A" is the correct answer. It certainly has been a lesson I've had to learn the hard way.

Have you heard of $ ?

bar = foo[0 .. $];

Additionally, try turning on warnings for your original code, and you'll notice you get one.

It probably is time to deprecate length inside array braces.

> Besides, is "foo[0 .. length]" really better than "foo[0 .. foo.length]"? The first is slightly more convenient but the second is more explicit and readable.

How about things like

int[] x = ALongClassName.AnotherClassName.aNonTrivialOperation()[1 .. $];

It's certainly nice to have a shortcut in cases like this, when typing out the original expression is not only prohibitively long, but could possibly waste performance re-evaluating the reference to the array.

Jarrett Billingsley wrote:
> Have you heard of $ ?
> 
> bar = foo[0 .. $];

I did notice the mention of "$" in the old newsgroup thread. However, I can't find any official documentation on it.


> Additionally, try turning on warnings for your original code, and you'll notice you get one.

Thanks for pointing out the warnings. I wasn't aware of that warning. Because libraries often cause many warnings, I don't regularly have warning turned on.


> How about things like
> 
> int[] x = ALongClassName.AnotherClassName.aNonTrivialOperation()[1 .. $];
> 
> It's certainly nice to have a shortcut in cases like this, when typing out the original expression is not only prohibitively long, but could possibly waste performance re-evaluating the reference to the array. 
> 

Why not the following?

int[] ref = ALongClassName.AnotherClassName.aNonTrivialOperation();
int[] x = ref[1 .. ref.length];

Or perhaps?

int[] x = ALongClassName.AnotherClassName.aNonTrivialOperation();
x = x[1 .. x.length];

Will use of "$" be faster than using a local reference?

Unless there is a reason not to create a local reference to an array, I don't think either "length" or "$" are necessary. At least "$" shouldn't cause programming errors, but it does make the language more complex.

Thanks,
  Bradley

Bradley Smith wrote:
> Jarrett Billingsley wrote:
> 
>> Have you heard of $ ?
>>
>> bar = foo[0 .. $];
> 
> 
> I did notice the mention of "$" in the old newsgroup thread. However, I can't find any official documentation on it.
> 
> 
>> Additionally, try turning on warnings for your original code, and you'll notice you get one.
> 
> 
> Thanks for pointing out the warnings. I wasn't aware of that warning. Because libraries often cause many warnings, I don't regularly have warning turned on.
> 
> 
>> How about things like
>>
>> int[] x = ALongClassName.AnotherClassName.aNonTrivialOperation()[1 .. $];
>>
>> It's certainly nice to have a shortcut in cases like this, when typing out the original expression is not only prohibitively long, but could possibly waste performance re-evaluating the reference to the array.
> 
> 
> Why not the following?
> 
> int[] ref = ALongClassName.AnotherClassName.aNonTrivialOperation();
> int[] x = ref[1 .. ref.length];
> 
> Or perhaps?
> 
> int[] x = ALongClassName.AnotherClassName.aNonTrivialOperation();
> x = x[1 .. x.length];

The original reason, IIRC, was with regard to using arrays + templates, where it's not always so easy to leverage temps. However, that orignal change brought us the maligned x[0..length] and its special cases. Took a long time (a year, perhaps?) and many frustrating battles to get that issue rectified, and the result was that sole '$' token

> 
> Will use of "$" be faster than using a local reference?
> 
> Unless there is a reason not to create a local reference to an array, I don't think either "length" or "$" are necessary. At least "$" shouldn't cause programming errors, but it does make the language more complex.

It was argued at the time that $len should be used, so that the $ prefix could thus signify meta-notions in general ($file, $line, $time, etc). In fact, it was perhaps the only time I recall a general concensus on anything ;)

But, that notion went nowhere as you can see. We just move on

"Bradley Smith" <digitalmars-com@baysmith.com> wrote in message news:e658a2$1r3s$1@digitaldaemon.com...

> Will use of "$" be faster than using a local reference?
>
> Unless there is a reason not to create a local reference to an array, I don't think either "length" or "$" are necessary. At least "$" shouldn't cause programming errors, but it does make the language more complex.

As Brad said, sometimes you don't have the luxury of using temporaries, such as with templates.

I doubt $ is any faster, performancewise, but it's certainly faster when typing.  I also wouldn't say that $ makes the language more complex - it adds a small amount of complexity, yes, but you also have to realize that it's only used in slice expressions, which in themselves are something that add complexity to the language!  I just think of $ as a feature of slice expressions.